The present invention relates to a storage system constituted of a plurality of storage sub-systems connected to computers via networks.
High reliability and high performance of a large scale storage system are required more and more nowadays. As a method of configuring a large scale storage system satisfying such user requirements, a plurality of storage sub-systems (hereinafter also called a “storage box”) are interconnected by an internal network and controlled independently. For example, as one practical implementation of a storage system, storage boxes are interconnected by a high speed private network and an I/O forward function is provided to forward an I/O request between storage boxes, so that a host can regard this system as one large scale storage system. If an I/O request received from a host is a request to another storage box, the I/O forward function communicates with the other storage box and returns a necessary message back to the host.
As a second method, storage boxes are interconnected not by a private network but by a general network via which an I/O request from a host is transmitted, to provide the I/O forward function.
iSCSI (Internet Small Computer System Interface) is expected as a protocol for configuring such a storage system, the specifications of iSCSI being stipulated by the IETF (Internet Engineering Task Force).
iSCSI is a protocol which performs an I/O process such as data storage by transferring a SCSI command and the like over a network whose communications are performed by the TCP/IP protocol. Ethernet (registered trademark), which is a typical network device capable of using the TCP/IP protocol, can be installed at lower cost than Fibre Channel presently used in a storage network. If iSCSI prevails, hosts and storage sub-systems larger in number than those of a storage network connected by Fibre Channel can be interconnected.
Patent Publication JP-A-2000-293317 discloses a method of detecting a load state of each part in a storage sub-system and avoiding load concentration.